Risk-Based Reliability Assessment of Modern Power Systems using Machine Learning and Probability Theory

Abstract

Risk-based reliability assessment is prevalent for modern power systems under higher penetration of renewable generations. This paper highlights the importance of machine learning and probabilistic approaches for risk-based reliability assessment during power system operation and planning. A set of metrics for realistic risk-based reliability assessment considering over-limit probabilities and corresponding severities is suggested. Probabilistic load flow using Monte-Carlo simulation is used to estimate the over-limit probabilities of power system variables. A detailed presentation of steps for the generation of random samples of a set of correlated random variables, development of realistic risk metrics, and portrayal of their significances via critical result analyses for different cases is expected to serve as a reference text for novice researchers in the field of risk-based reliability assessment of modern power systems integrated with photovoltaic generations.